Message/Author |
|
Hoon Lee posted on Tuesday, August 07, 2012 - 8:31 am
|
|
|
Dear Dr. Muthen, I am running a moderated mediation model, which includes one mediator and two moderators. I am wondering if I can obtain bias-correct CIs for inidrect effects at different levels of moderators. Just in case, I am attaching the Mplus code that I am currently using. Thank you very much in advance. DATA: FILE = M:\2008 2010 mplus.dat; FORMAT is FREE; VARIABLE: names = age gen edu inc pin pst npf tvf onl tal mhe mpr par cro ief; usevariables = age gen edu inc pin pst npf tvf onl tal mhe mpr par cro ief mv xw; missing = all(999); DEFINE: mv = cro*ief; xw = mhe*mpr; ANALYSIS: !type = basic; bootstrap = 5000; MODEL: par on cro(b1) age gen edu inc pin pst npf tvf onl tal mhe mpr ief mv(b3) xw; cro on mhe(a1) age gen edu inc pin pst npf tvf onl tal mpr xw(a3); MODEL CONSTRAINT: new (ind wmodval vmodval); wmodval = 1; vmodval = 1; ind=(a1+a3*wmodval)*(b1+b3*vmodval); output: cinterval (bcbootstrap); sampstat standardized; |
|
|
I don't believe we give confidence intervals for NEW parameters in MODEL CONSTRAINT. You should try it out to be sure. |
|
Hoon Lee posted on Tuesday, August 07, 2012 - 2:37 pm
|
|
|
Thank you so much for your prompt response. I tried out the above syntax and found that it produced a bc CI for the new paramenter ("ind"). However, it only produced an unstrandardized CI for that, and I am still wondering if there is a way to command Mplus to produce a standardized CI. Thank you very much again! |
|
|
There is no option to expand the CINTERVAL option. Everything available is given. |
|
|
You would have to define the standardized indirect effect yourself in Model Constraint. |
|
Back to top |